Learn

Practical articles for developers who build AI agents or have to read, draw or document code they did not write. Each one is backed by what we built and measured in Nehemvix.

Agentic AI engineering

• Agentic AI examples: one production agent, measured: what makes an AI system agentic, shown on one real agent with its graph, its pauses and how it is measured.
• Human in the loop for AI agents: where an agent should stop and wait for a person, and where it can safely carry on alone.
• Pause and resume an agentic AI pipeline without paying twice: how a long LangGraph job survives a quota, a crash or a model switch without redoing finished calls.
• Rate limits, retries and switching models in LLM apps: how to tell "wait a minute" from "out for today" or "wrong key", and what to do about each.
• Estimate and cap LLM costs before a run: how we estimate a job before spending anything, hold a cap when calls run in parallel, and how close the estimates came.

Evaluating and grounding AI

• Catch hallucinated function and file names in AI-written docs: how to check every name a model writes against the code, and turn the result into a score.
• Build an eval gate for an LLM application: how to measure a checker on hand-checked guides and blind rounds, and stop a release when a number slips.
• Test for prompt injection from repository content: how we attack our own pipeline with a hostile repository, live on three providers.

AI repository analysis

• How AI reads a large repository: map, reduce, synthesize, write: how a model can understand thousands of files while reading each one only once.
• Send code to an LLM safely: what to remove, pin and refuse before any of your code reaches a model.

Agents and protocols

• MCP server vs A2A: connect AI assistants to your own data: what each protocol is for, and what an outside assistant should and should not be allowed to do.
• How to understand an agentic AI codebase quickly: where to find the agent loop, its tools, its memory and its guardrails in someone else's code.
• How to document a LangGraph project: how to find the graph, its state, its interrupts and its checkpoints, and check each claim in the code.

Reading and documenting code

• How to understand a new codebase: a method you can follow by hand, from the README to tracing one request.
• Draw a system architecture diagram from your code: why hand-drawn diagrams drift, and what a diagram drawn from the code can and cannot show.
• Nehemvix vs DeepWiki: a factual comparison, as of November 2026, of two ways to turn a repository into documentation.

See it applied

How Nehemvix turns a repository into a guide and how every guide is checked. The full documentation covers each feature step by step.

For the practice beyond one tool, on any stack, we publish AELSHA™ Insights on building, running and securing agents.

Stuck on one of these in your own code? Ask AELSHA™ for an expert session.